phc-discussions - Re: [PHC] An argument for Catena-2

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CAOLP8p6CkRdqVzL4efSAkzgrPHkXR3s4bHPHy8Oh5NpeA_V13Q@mail.gmail.com>
Date: Mon, 27 Jan 2014 23:00:26 -0500
From: Bill Cox <waywardgeek@...il.com>
To: discussions@...sword-hashing.net
Subject: Re: [PHC] An argument for Catena-2

Yep.  This is all single threaded, and the comparison is to the
default Script package in Arch Linux on my son's MineCraft server (my
"development" machine).  Check out the waywardgeek branch of Catena:

https://github.com/cforler/catena/tree/waywardgeek

Christian was kind enough to let me have commit rights, but I have a
habit of breaking code, so I really have to have my own sandbox.  I
created the waywardgeek branch and have had a great time hacking it.

I replaced the hashing with the same multiply bound hash in NoelKDF,
and also changed H_LEN to 4096.  That's all it took to speed it up
dramatically.  You can checkout catena from:

 https://github.com/cforler/catena

These results should be easily reproducible.

Bill

On Mon, Jan 27, 2014 at 10:57 PM, Solar Designer <solar@...nwall.com> wrote:
> Bill,
>
> On Tue, Jan 28, 2014 at 07:33:43AM +0400, Solar Designer wrote:
>> On Sun, Jan 26, 2014 at 03:32:04PM -0500, Bill Cox wrote:
>> > I benchmarked Catena-2, and it is running 3.4X faster than Scrypt,
>> > compared to 2.6X faster for Catena-3.
>>
>> Why are Catena-2 and -3 faster than scrypt, in your benchmarks?
>
> Oh, I guess you meant Catena with "the NoelKDF multiply hash", not one
> of Christian's versions.  Then this could be so, for some thread count
> to memory bandwidth ratios.
>
> Alexander